Scientists at IBM’s T.J. Watson Research Center in Yorktown Heights, N.Y., have carefully measured the slowing of light in a silicon photonic crystal and have demonstrated that the amount of slowing can be controlled by heating the crystal. Slow light in photonic crystals previously had been demonstrated by the IBM researchers and others, but implementing the effect in a millimeter-size piece of silicon fabricated with conventional techniques, making careful interferometric measurements of the effect, and tuning the degree to which light is slowed are important advances toward the practical application of the phenomenon.Figure 1. The input side of the Mach-Zehnder interferometer comprised a 15° Y-splitter in a silicon strip waveguide (top). The strip waveguide was butt-coupled to the photonic-crystal waveguide fabricated in a 223-nm-thick silicon membrane (bottom). To unbalance the interferometer — that is, to slow light more in one arm than in the other — the scientists fabricated the crystals in the two arms with slightly different hole sizes. Images ©Nature Publishing Group.To measure the slowing of light in the photonic-crystal waveguide, the researchers fabricated a Mach-Zehnder interferometer in which light was slowed significantly in one arm but only marginally in the other (Figure 1). When the light in the two arms was combined at the output end of the interferometer, the wavelength-dependent interference pattern was the result of wavelength-dependent slowing of light in one arm (Figure 2).Figure 2. The transmission through the interferometer showed a periodic variation that was caused by the slowing of light in one arm. The farthest-left point on the red trace (~1500 nm) corresponds to destructive interference between the two arms. When the incident wavelength increased to ~1520 nm, light in one arm of the interferometer was slowed sufficiently so that constructive interference occurred when the arms were combined. As the incident wavelength increased further, the period of the interference pattern decreased, indicating that light in the slow arm was even slower. As indicated by the blue traces, the period of the interference pattern decreased dramatically with increasing wavelength. The scientists calculated that the smallest period they could reliably measure corresponds to a slowing of light by more than a factor of 300 from its velocity in space; i.e., from 3 × 108 m/s to less than 1 × 106 m/s.One application of slow light is optical buffering; that is, storing optical data for short periods. Buffering is important, for example, to ensure that same-wavelength signals passing through an optical router do not overlap. If the photonic crystals demonstrated at T.J. Watson are viable for this application, they must have sufficiently low chromatic dispersion and propagation loss.The researchers estimate that the chromatic dispersion of their crystals is 500 ps/nm/mm, a value ~107 times greater than that of conventional telecommunications fiber. But because the length of a practical photonic-crystal buffer is on the order of a millimeter, its total dispersion nonetheless may be within acceptable limits.Propagation loss through the photonic crystal is negligible, but a large loss is introduced at the interface between the crystal and the linking waveguide. Further work on this interface is needed, the investigators believe, if these devices are to become effective optical buffers.Figure 3. When a 1-MHz square wave was applied to the heating element, the interferometer switched from opaque to transparent in 100 ns.Variable optical buffers, as well as dynamic dispersion compensators, can be fabricated if the degree to which light is slowed can be tuned with an electrical signal. The scientists demonstrated this by heating the photonic crystal, changing its refractive index through the thermo-optic effect. To minimize the loss introduced by the heating elements, they placed them beside, rather than above and below, the waveguide. When they drove the heaters with a 1-MHz square wave, they observed that the transmission through the interferometer switched from opaque to transparent in only 100 ns (Figure 3). They believe that this is the fastest response ever reported for a thermo-optic modulator.